The mystery & magic of skin tones?

mochi

I've been lurking here and elsewhere and learning as much as I can about skin tones. I've read Lee Varis's book, I've read a lot of Dan Margulis, and I've read lots of posts by Andrew Rodney (though I haven't read his book). My head is swimming with great information, but I'm hoping to condense it into a working method that is relatively consistent and works.

It seems to me so far that the single most important factor in achieving pleasing skin tones is proper WB. When the WB is off, I find that it's very hard to get good skin tones no matter what I do. When the WB is right on, it seems that only minor adjustments are necessary to get good results. I've started using a WhiBal card, especially in challenging light, and that seems to help a lot.

But there are inevitably times I forget to use the card or when it's not practical. And sometimes during post-processing it can be difficult to find something neutral in the scene to use for WB adjustment. In these cases I have a hard time getting a good WB to begin with for color correction, and although I'm learning I still don't trust my eyes completely. (i.e. something looks good to me, then I submit it for critique, changes are made, and it looks much better. Why didn't I notice that?)

So, a few questions:

1. Any tips on getting the WB right when there's nothing necessarily neutral in the scene?

2. What color space should I correct for skin tones in? I've begun using LAB because of its simplicity, but I'm uncertain about this area.

3. I have a skin tone chart that lists light caucasian skin at 66-14-17. Obviously that's not an absolute, but how much do we strive towards the numbers and how much do we depend on our eyes?

Here are two images as an example. The first was just adjusted for WB in Lightroom. The second was corrected in Photoshop with LAB color space. I placed a sampler on her forehead and moved the top-right point on the curve in Lightness, A & B channels to get 66-14-17.

Which do you prefer?

ziggy53

Mochi,

Welcome to the Digital Grin.

I've had great results using the technique described here:

http://www.smugmug.com/help/skin-tone

I do sometimes have to try a few different sample points in order to find one that works well to adjust against.

You are correct about white balance and if you have a mixture of different kinds of light it can be next to impossible to achieve good/pleasing/proper skin tones.

Since most modern photographic CCD and CMOS sensors (and derivatives) are designed to balance best against daylight, let's say 5200K just for this argument, it makes sense to try to shoot in that white balance, as any other WB will probably digitally expand/interpolate tones in order to achieve color balance.

Electronic flash, balanced for daylight, seems to be the preferred source for approximating daylight in a controlled environment (studio i.e.), and that is my personal preference as well.

So, relating to your first question, I do try to create daylight as much as possible if the ambient is otherwise. This means using flash and, sometimes, multiple flash. Assuming I am in control of the lighting, I can use daylight WB predictably on the camera and get reasonably consistant results in post-production.

If shooting in available light only, perhaps a church during the ceremony or casual home photos, I tend to just use the above linked technique for the flash tones and not worry too much about the absolute color purity of clothing and background.

Sometimes, like for a longish gymnasium shot, there is little you can do but accept the available light, no matter how poor. If using flash is not a practical option, I just shoot in RAW and then hope to be able to adjust using any number of techniques, but accepting anything "close" to reality.

A couple of years ago, I was shooting a college graduation and a very popular graduate, who happened to be blind, received her diploma. I was astonished to see how different her skin tones were from "normal" Caucasian skin tones (possibly from little exposure to sunlight, but I really don't know.)

I was determined to create an acceptable tonality, despite the horrible gym lighting and her natural tonal differences. I found that a combination of corrections yielded good results, but that judicious "Channel Mixer" against just certain colors in CMYK color space was what worked best overall and that correction just against her visible skin (using a mask and layer). I doubt that you will find any sort of exacting "formulae" in any printed or on-line documentation for that problem, but determination is what solved the problem for me.

My point, if any, is that no single technique covers all circumstances. Find a number of different techniques and use whatever seems appropriate from the sources you mentioned as well as anything else you might try.

As for the two images you posted, I guess I would like something in-between the two, but more like the first as I think there is more warmth in the first and that probably matches the mood of the scene a little better IMO.

arodney

mochi wrote:

It seems to me so far that the single most important factor in achieving pleasing skin tones is proper WB. When the WB is off, I find that it's very hard to get good skin tones no matter what I do. When the WB is right on, it seems that only minor adjustments are necessary to get good results. I've started using a WhiBal card, especially in challenging light, and that seems to help a lot.

Yup, its pretty critical. And even with scenes shot under non standard illuminates (mid day sunlight), you can do a lot to improve the image with proper WB. And yes, you want to do this in the Raw processor way before you render the pixels as colors! Far more control, faster etc.

1. Any tips on getting the WB right when there's nothing necessarily neutral in the scene?

With Raw WB, you want a non specular white if possible. But you can really hose the color appearance if you WB in the wrong area. Sometimes, it doesn't work well, so I just use the Temp and Tint sliders in ACR or LR and adjust visually. Of course, having a small WB card or mini MacBeth makes this so much easier.

2. What color space should I correct for skin tones in? I've begun using LAB because of its simplicity, but I'm uncertain about this area.

The Color Space of whatever Raw converter you're using (in the case of LR and ACR, that's ProPhoto RGB). But you don't need to be overly concerned here. The numeric values for what someone might suggest is "proper" skin tones will vary based on the color space so again, doing all this at the Raw rendering stage makes this so much easier as you only have to "learn" one set of numbers, the color space the processor is using. Even with LR, where you're getting percentages, not RGB values, the key is looking at successful rendering you've done and getting comfortable with the numbers.

3. I have a skin tone chart that lists light caucasian skin at 66-14-17. Obviously that's not an absolute, but how much do we strive towards the numbers and how much do we depend on our eyes?

Assuming you have a calibrated display and you're not color blind, trust your eyes first! And the values above will totally vary depending on the color space. So if you're using differing color spaces in Photoshop, it just makes this all the more difficult. Once again everyone, do this at the Raw conversion stage.

Of the two images, at least on the web, I prefer the top version. Yes, overall, there's a warm cast to the image but it feels far more like what I suspect the scene looked like and cool skin is usually not as pleasing as warm. But which do YOU prefer? Not only is the 2nd image cooler, it looks a bit "down".

mochi

ziggy53 wrote:

Mochi,

As for the two images you posted, I guess I would like something in-between the two, but more like the first as I think there is more warmth in the first and that probably matches the mood of the scene a little better IMO.

Thanks for your reply, ziggy53! I have seen the smug mug tutorial - in fact, that's one of the first sources of information I found on the subject.

I agree with you that something in between the first and second image would probably be ideal, though I'm leaning closer to the first at this second viewing.

mochi

arodney wrote:

With Raw WB, you want a non specular white if possible. But you can really hose the color appearance if you WB in the wrong area. Sometimes, it doesn't work well, so I just use the Temp and Tint sliders in ACR or LR and adjust visually. Of course, having a small WB card or mini MacBeth makes this so much easier.

Thanks for your reply, Andrew. I have a WhiBal card I've been trying to get in the habit of using. Also, I'm starting to use both on-camera flash with a diffuser/reflector and off-camera flash much more often to bring the color temp closer to daylight since the 5D WB seems to be so much better in that range.

arodney wrote:

The Color Space of whatever Raw converter you're using (in the case of LR and ACR, that's ProPhoto RGB). But you don't need to be overly concerned here. The numeric values for what someone might suggest is "proper" skin tones will vary based on the color space so again, doing all this at the Raw rendering stage makes this so much easier as you only have to "learn" one set of numbers, the color space the processor is using. Even with LR, where you're getting percentages, not RGB values, the key is looking at successful rendering you've done and getting comfortable with the numbers.

I'd really love to learn to color correct in LR with the percentages. I guess I just need to work out what those percentage numbers should be with caucasian skin tones. On another forum I saw a suggestion that red should be 112.3% of green, and that blue should be 97.7% of green. I'm not sure how tha author of that post came up with those numbers. I'd love to hear your thoughts on this, because if I could correct for skin in LR that would be great!

I understand (at least I think I do) that if I want more blue I move the WB slider left; If I need more red I move the tint slider towards magenta, and for more green I move the tint slider the other way. Is that how you do it in LR/ACR?

Of course there are also the HSL and RAW calibration sliders, but I want to get as comfortable as I can with nailing WB before I introduce several other variables.

That said, I did calibrate my 5D to ACR using Rags Gardner's Photoshop script. I ran a separate calibration with the skin tone patch in place of the red one, and that seems to provide a good starting place for portraits

arodney wrote:

Assuming you have a calibrated display and you're not color blind, trust your eyes first! And the values above will totally vary depending on the color space. So if you're using differing color spaces in Photoshop, it just makes this all the more difficult. Once again everyone, do this at the Raw conversion stage.

I guess the trick right now is that I don't always trust my eyes. I find that when I do several iterations of a particular image, it becomes more difficult for me to hone in on what I like. Leaving the computer for a while and coming back often helps. That's why I'd at least like a fair understanding of the numbers and how they work; then I can choose to ignore them if that's what looks best to me.

I always use Lightroom (ppRGB) and Photoshop CS3 (ppRGB). I calibrate weekly with an i1 Display 2.

arodney wrote:

Of the two images, at least on the web, I prefer the top version. Yes, overall, there's a warm cast to the image but it feels far more like what I suspect the scene looked like and cool skin is usually not as pleasing as warm. But which do YOU prefer? Not only is the 2nd image cooler, it looks a bit "down".

arodney

Build a group of successful images of skin tone's (one's you output or viewed on the web) even those that are rendered. Import into Lightroom and read the skin tones. The values are based on the internal color space of this product. You'll soon find values (probably too many) that you'll get comfortable with that will work within LR. Frankly, I find the by the numbers approach to work really well with some aspects of imagery (highlights, neutrals, shadows) but skin tone is all over the map. If you ever get a Spectrophotometer like an EyeOne Pro, its fun to actually measure skin tones and see how they vary. And this is all based on one light source! I think you'll find you will end up doing a much better job trusting your eyes with that calibrated display. Of course, a calibrated display AND numbers are hard to beat. You'll find they both back each other up.

In the end, a lot of this is subjective.

pathfinder

What color was the wall behind her head?

Using the WB eye dropper on the wall in the upper right corner, in ARC with the first jpg, gives temperature of -16 and a tint of -5 and looks promising. Maybe needs just a touch more magenta.

Did you balance off the wall in the second image?

I think the images are just a little dark also - maybe 1/3 to 1/2 stop.

Duffy Pratt

If you are in Photoshop, you can always use the info panel to get the numbers from a different color space than the one you are working in. I very often have the number readings as the [working space number] and [LAB numbers].

If you are going to make the adjustments in Lightroom, then you will probably need to convert the LAB numbers you are comfortable with to the LR %'s. This shouldn't be that hard to do. Just take some pictures that you think look right, get their LR numbers, and then open in PS and get the LAB numbers.

One of the big tricks with judging skin by the numbers is that the expected colors should change with the lighting levels. The brighter the lighting, the more likely the color numbers will bias toward white. Go dark enough, and the colors, I think, will start going neutral again.

Duffy

pathfinder

The other factor in using numbers for skin, is that you CANNOT use females cheeks or chins - they have almost uniformly been covered with makeup.

The edge of the forehead near the hairline, backs of hands or forearms, the sides of necks sometimes. You have to be a detective sometimes.

Sometimes the WB eye dropper will work rather well on the white sclera of the eye - sometimes it is just too blue. Sometimes you can get a WB off an incisor. Or a white pearl necklace. Or a white T-shirt. Gray hair works sometimes.

mochi

Thanks everyone for your very helpful replies. I like the idea of choosing some "successful" images and averaging the RGB percentages in LR.

In the photo on the bottom I believe I used the WB dropper on the sclera of her eye. It's probably too blue, though, and perhaps that's why the image lost a lot of its warmth.

I feel like I have a lot of good things to try out now. Thanks again, and I'll report back!

arodney

Duffy Pratt wrote:

If you are going to make the adjustments in Lightroom, then you will probably need to convert the LAB numbers you are comfortable with to the LR %'s. This shouldn't be that hard to do. Just take some pictures that you think look right, get their LR numbers, and then open in PS and get the LAB numbers.

If you know the correct LR values, why on earth jump to Photoshop and mess with Lab?

Seriously, this Lab addiction around here is alarming. Talk about a non intuitive color model and one that's totally unnecessary in a product like LR (one you're going to start using in the workflow).

Duffy Pratt

Andrew,

The suggestion was to find LR values, since he does not already know those values and is comfortable to some extent with LAB.

Duffy

arodney

Duffy Pratt wrote:

Andrew,

The suggestion was to find LR values, since he does not already know those values and is comfortable to some extent with LAB.

Duffy

Just take some pictures that you think look right, get their LR numbers, and then open in PS and get the LAB numbers.

If you take numbers that look right (and output right), you've got the numbers. What good is it to now translate into another color space that the current application doesn't support (nor need), especially when the current application (LR) is where you'll be correcting for skin tone?